Abstract: With considerations of the uncertainties of damage such as randomness and fuzziness, the overall instantaneous stiffness matrix of the structure is established, and the matrix similarity is applied to measure the resemblance between the initial matrix and the damaged ones. On this basis, the stiffness damage index is established, considering the importance of different parts of the bridge. Then, a group of fuzzy sets representing standard damage states could be inferred with the normal membership function, and the maximum membership principle is used to identify the degree of structural damage. The proposed method is further demonstrated through a two-span bridge model and the results show that the fuzzy identification of damage degree could be achieved effectively without regard to the state similarity matrix and the probability distribution of the structural damage index. The method has the potential to be embedded into the Structural Health Monitoring System, and to give decision support to the structural maintenance and life prediction.